How to generate an AI video from a text prompt — with sound, up to 15 seconds per clip
Updated 10 September 2026 · Twin AI Labs
Type a prompt, pick one of 19 video models — Veo 3.1, Kling 3.0, Seedance 2.0, Hailuo 2.3, Wan 2.7 — and Twin AI renders a 5–15-second clip from 600 credits. Veo 3.1 generates audio built-in, Kling 3.0, Kling 2.6 and Seedance 1.5 add sound as an option; a 30-second video is three 10-second clips joined together.
How to generate a video from text: 6 steps
The panel on the left is the tool itself. An account is only asked for when you press Generate — your prompt and uploaded frames are kept, and a new account starts with free credits, no card.
- Pick a model. The “Model” button opens a drawer with all 19 video models, their credit price and a preview; the choice decides which parameter chips appear below.
- Write the prompt: subject, action, camera movement, lighting. Short concrete sentences beat a paragraph; for Veo you can add an audio cue such as “audio: crowd noise”.
- Optionally upload a start frame (and, on models with two slots, an end frame). With a photo attached the model works in image-to-video mode and animates your picture.
- Set duration, resolution, aspect ratio and sound with the parameter chips. Each option shows its credit delta and the total updates on the Generate button.
- Press Generate. The clip lands in the gallery on the right; the API estimates 1.5 to 4 minutes per render.
- Open the clip and press Download to save the MP4 — or publish it to the feed and share the link.
How to get a 30-second video with sound
No model on this page renders 30 seconds in one pass: the ceiling is 15 seconds on Wan 2.6 (no sound) and 12 seconds on Seedance 1.5 (sound optional). The practical route is three 10-second clips from the same model with a consistent scene description, joined in any video editor (CapCut, iMovie, DaVinci Resolve). To keep continuity, take the last frame of clip one and upload it as the start frame of clip two.
The math at API prices. Kling 3.0 with sound: 3 × (600 + 220 for 10 seconds + 100 for sound) = 2,760 credits for 30 seconds. Seedance 1.5 with sound: two 12-second clips and one 8-second clip — 2 × (600 + 40 + 20) + (600 + 20 + 20) = 1,960 credits for 32 seconds, and 1080p carries no surcharge there. Veo 3.1 generates audio built-in but exposes no duration setting in the composer — 900 credits per clip.
Video models available on this page
All 19 models are live in the composer right now. Elo is the Twin AI leaderboard rating from user side-by-side votes; the price is the base cost of a clip at default settings, and “+N” is the credit delta for that option.
| Model | Elo | Credits per clip | Duration | Sound | Resolution | Input |
|---|---|---|---|---|---|---|
| Seedance 2.0 | 1272 | 600 (Standard +140) | 5 s / 10 s (+650) | — | 480p / 720p / 1080p (+1500) | text or image |
| Kling 3.0 | 1251 | 600 (Pro +240) | 5 s / 10 s (+220) | option, +100 | — | text or image |
| Grok Imagine | 1233 | 600 | 6 s / 10 s (+20) | — | 480p / 720p | text or image |
| Kling Turbo | 1210 | 600 | 5 s / 10 s (+60) | — | — | text or image |
| Veo 3.1 | 1209 | 900 | — | built-in | — | text or image |
| Veo 3.1 Fast | 1208 | 600 | — | built-in | — | text or image |
| Wan 2.6 | 1192 | 600 | 5 s / 10 s (+210) / 15 s (+460) | — | 720p / 1080p (+300) | text or image |
| Kling 2.6 | 1192 | 600 | 5 s / 10 s (+490) | option, +100 | — | text or image |
| Kling Pro | 1188 | 600 | 5 s / 10 s (+70) | — | — | image only |
| Hailuo 2.3 | 1178 | 600 (Pro +20) | 6 s / 10 s (+30) | — | 768p / 1080p (+30) | image only |
| Hailuo | 1177 | 600 (Pro +30) | 6 s / 10 s (+30) | — | 768p / 1080p (+10) | text or image |
| Kling Standard | 1176 | 600 | 5 s / 10 s (+40) | — | — | image only |
| Seedance 1.5 | 1174 | 600 | 4 s / 8 s (+20) / 12 s (+40) | option, +20 | 480p / 720p / 1080p | text or image |
| Kling Motion Control | 1142 | 600 | — | — | 720p / 1080p (+150) | image + video |
| Kling Master | 1139 | 600 | 5 s / 10 s (+650) | — | — | text or image |
| Bytedance Fast | 1135 | 600 | 5 s / 10 s (+30) | — | 720p / 1080p (+50) | image only |
| Wan Turbo | 1111 | 600 | — | — | 480p / 720p (+60) | text or image |
| Wan 2.7 | — | 600 | 5 s / 10 s (+300) | — | 720p / 1080p (+170) | text or image |
| HappyHorse | — | 600 | 5 s / 10 s (+510) | — | 720p / 1080p (+250) | text or image |
Other ways to generate a video
If you want one specific model from its own vendor, these are the three tools people usually name. Here is the neutral view, then what is different about Twin AI.
- Runway — its own Gen-4 models for text-to-video and image-to-video in the Runway web app, on subscription plans.
- Pika — text-to-video and video effects in the Pika app, with a developer API at dev.pika.art.
- Luma Dream Machine — Luma AI’s web app for generating video from text and from photos.
- What is different in Twin AI: 19 video models — Veo 3.1, Kling 3.0, Seedance 2.0, Hailuo, Wan — and 30+ models overall on one balance, credits billed only for what you generate, no per-vendor subscriptions.
Sources
- Twin AI public model list (prices, parameters)api.twinailabs.com
- Runwayrunway.com
- Pika — developer platformdev.pika.art
- Luma AI — Dream Machinelumalabs.ai
FAQ
Can I generate a 30-second AI video with sound from a text prompt?
Not in one pass — the longest clip on Twin AI is 15 seconds (Wan 2.6) and the longest with a sound option is 12 seconds (Seedance 1.5). Generate three 10-second clips with Kling 3.0 and sound on (+100 credits each) and join them; total 2,760 credits.
Which models add sound?
Veo 3.1 and Veo 3.1 Fast generate audio built-in. Kling 3.0 and Kling 2.6 have a “With Sound” option for +100 credits, Seedance 1.5 has an “Audio” option for +20. The other models on the page expose no sound switch in the composer.
How much does one AI video cost?
From 600 credits for a 5-second clip on most models; Veo 3.1 is 900. Longer duration, 1080p, Pro quality and sound add the deltas listed in the table, and the Generate button always shows the final price before you confirm.
Can I animate my own photo?
Yes. Upload it as the start frame: all 19 models support image-to-video, and Kling Pro, Kling Standard, Kling Motion Control, Bytedance Fast and Hailuo 2.3 work only from an image.
What resolution and aspect ratio can I get?
1080p on Wan 2.7, Wan 2.6, Seedance 2.0, Seedance 1.5, HappyHorse, Bytedance Fast, Hailuo, Hailuo 2.3 and Kling Motion Control; 720p or 768p elsewhere, or no resolution setting at all. Aspect ratios 9:16, 16:9 and 1:1 on most models, 21:9 on Seedance.